Face Generation¶

In this project, you'll use generative adversarial networks to generate new images of faces.

Get the Data¶

You'll be using two datasets in this project:

MNIST
CelebA

Since the celebA dataset is complex and you're doing GANs in a project for the first time, we want you to test your neural network on MNIST before CelebA. Running the GANs on MNIST will allow you to see how well your model trains sooner.

If you're using FloydHub, set data_dir to "/input" and use the FloydHub data ID "R5KrjnANiKVhLWAkpXhNBe".

#data_dir = './data'

#FloydHub - Use with data ID "R5KrjnANiKVhLWAkpXhNBe"
data_dir = '/input/R5KrjnANiKVhLWAkpXhNBe'


"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import helper

helper.download_extract('mnist', data_dir)
helper.download_extract('celeba', data_dir)

Downloading mnist: 9.92MB [00:01, 7.48MB/s]                            
Extracting mnist: 100%|██████████| 60.0K/60.0K [00:11<00:00, 5.23KFile/s]
Downloading celeba: 1.44GB [00:22, 63.4MB/s]

Extracting celeba...

Explore the Data¶

MNIST¶

As you're aware, the MNIST dataset contains images of handwritten digits. You can view the first number of examples by changing show_n_images.

show_n_images = 25

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
%matplotlib inline
import os
from glob import glob
from matplotlib import pyplot

mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'mnist/*.jpg'))[:show_n_images], 28, 28, 'L')
pyplot.imshow(helper.images_square_grid(mnist_images, 'L'), cmap='gray')

<matplotlib.image.AxesImage at 0x7feefa683f98>

CelebA¶

The CelebFaces Attributes Dataset (CelebA) dataset contains over 200,000 celebrity images with annotations. Since you're going to be generating faces, you won't need the annotations. You can view the first number of examples by changing show_n_images.

show_n_images = 25

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'img_align_celeba/*.jpg'))[:show_n_images], 28, 28, 'RGB')
pyplot.imshow(helper.images_square_grid(mnist_images, 'RGB'))

<matplotlib.image.AxesImage at 0x7feef0417a58>

Preprocess the Data¶

Since the project's main focus is on building the GANs, we'll preprocess the data for you. The values of the MNIST and CelebA dataset will be in the range of -0.5 to 0.5 of 28x28 dimensional images. The CelebA images will be cropped to remove parts of the image that don't include a face, then resized down to 28x28.

The MNIST images are black and white images with a single color channel while the CelebA images have 3 color channels (RGB color channel).

Build the Neural Network¶

You'll build the components necessary to build a GANs by implementing the following functions below:

model_inputs
discriminator
generator
model_loss
model_opt
train

Check the Version of TensorFlow and Access to GPU¶

This will check to make sure you have the correct version of TensorFlow and access to a GPU

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
from distutils.version import LooseVersion
import warnings
import tensorflow as tf

# Check TensorFlow Version
assert LooseVersion(tf.__version__) >= LooseVersion('1.0'), 'Please use TensorFlow version 1.0 or newer.  You are using {}'.format(tf.__version__)
print('TensorFlow Version: {}'.format(tf.__version__))

# Check for a GPU
if not tf.test.gpu_device_name():
    warnings.warn('No GPU found. Please use a GPU to train your neural network.')
else:
    print('Default GPU Device: {}'.format(tf.test.gpu_device_name()))

TensorFlow Version: 1.1.0
Default GPU Device: /gpu:0

Input¶

Implement the model_inputs function to create TF Placeholders for the Neural Network. It should create the following placeholders:

Real input images placeholder with rank 4 using image_width, image_height, and image_channels.
Z input placeholder with rank 2 using z_dim.
Learning rate placeholder with rank 0.

Return the placeholders in the following the tuple (tensor of real input images, tensor of z data)

import problem_unittests as tests

def model_inputs(image_width, image_height, image_channels, z_dim):
    """
    Create the model inputs
    :param image_width: The input image width
    :param image_height: The input image height
    :param image_channels: The number of image channels
    :param z_dim: The dimension of Z
    :return: Tuple of (tensor of real input images, tensor of z data, learning rate)
    """
    # TODO: Implement Function
    
    image_shape = (image_width, image_height, image_channels)
    input_real = tf.placeholder(tf.float32, (None, *image_shape), name='input_real')
    input_z = tf.placeholder(tf.float32, (None, z_dim), name='input_z')
    lr = tf.placeholder(tf.float32, name='learning_rate')

    return (input_real, input_z, lr)


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_inputs(model_inputs)

Tests Passed

Discriminator¶

Implement discriminator to create a discriminator neural network that discriminates on images. This function should be able to reuse the variables in the neural network. Use tf.variable_scope with a scope name of "discriminator" to allow the variables to be reused. The function should return a tuple of (tensor output of the discriminator, tensor logits of the discriminator).

def discriminator(images, reuse=False):
    """
    Create the discriminator network
    :param images: Tensor of input image(s)
    :param reuse: Boolean if the weights should be reused
    :return: Tuple of (tensor output of the discriminator, tensor logits of the discriminator)
    """
    
    # TODO: Implement Function
    
    alpha = 0.1
    dropout = 0.3
    initializer = tf.contrib.layers.xavier_initializer()

    def conv2d(x, out_dim, k=5, s=2, padding='same', use_batch_norm=True, training=True):
        layer = tf.layers.conv2d(x, out_dim, k, s, padding, kernel_initializer=initializer)
        if use_batch_norm == True:
            layer = tf.layers.batch_normalization(layer, training=training)
        #layer = tf.layers.dropout(layer, rate=dropout, training=True)
        layer = tf.maximum(alpha*layer, layer)
        return layer

    with tf.variable_scope('discriminator', reuse=reuse):
        layer_1 = conv2d(images,   32, use_batch_norm=False, training=False) # => 14x14x32                       
        layer_2 = conv2d(layer_1,  64)  # => 7x7x64
        layer_3 = conv2d(layer_2, 128)  # => 4x4x128 
        layer_4 = conv2d(layer_3, 256)  # => 2x2x256

        flat = tf.reshape(layer_4, (-1, 2*2*256))

        logits = tf.layers.dense(flat, 1)
        output = tf.sigmoid(logits)

        return output, logits

    
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_discriminator(discriminator, tf)

Tests Passed

Generator¶

Implement generator to generate an image using z. This function should be able to reuse the variables in the neural network. Use tf.variable_scope with a scope name of "generator" to allow the variables to be reused. The function should return the generated 28 x 28 x out_channel_dim images.

def generator(z, out_channel_dim, is_train=True):
    """
    Create the generator network
    :param z: Input z
    :param out_channel_dim: The number of channels in the output image
    :param is_train: Boolean if generator is being used for training
    :return: The tensor output of the generator
    """

    # TODO: Implement Function
    
    alpha = 0.1
    dropout = 0.3
    initializer = tf.contrib.layers.xavier_initializer()

    def fully_connect(x, dims, training):
        layer = tf.layers.dense(x, dims[0]*dims[1]*dims[2], kernel_initializer=initializer)
        layer = tf.reshape(layer, (-1, dims[0], dims[1], dims[2]))
        layer = tf.layers.batch_normalization(layer, training=training)
        layer = tf.maximum(alpha*layer, layer)
        return layer

    def conv2d_transpose(x, out_dim, k=5, s=2, padding='same', 
                         use_batch_norm=True, use_leaky_relu=True, training=True):
        
        layer = tf.layers.conv2d_transpose(x, out_dim, k, s, padding, 
                                           kernel_initializer=initializer)
        if use_batch_norm == True:
            layer = tf.layers.batch_normalization(layer, training=training)
        #layer = tf.layers.dropout(layer, rate=dropout, training=True)
        if use_leaky_relu == True:
            layer = tf.maximum(alpha*layer, layer)
        
        return layer

    with tf.variable_scope('generator', reuse=not is_train):
        layer_1 = fully_connect(z, (4, 4, 512), is_train)           # => 4x4x512
        layer_2 = conv2d_transpose(layer_1, 128, k=4, s=1, 
                                   padding='valid',
                                   training=is_train)               # => 7x7x128  
        layer_3 = conv2d_transpose(layer_2, 64, training=is_train)  # => 14x14x64
        layer_4 = conv2d_transpose(layer_3, 32, training=is_train)  # => 28x28x32 

        logits = conv2d_transpose(layer_4, out_channel_dim, 3, 1, 
                                  'same', False, False, True)       # => 28x28x3(1)

        output = tf.tanh(logits)

        return output
        

"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_generator(generator, tf)

Tests Passed

Loss¶

Implement model_loss to build the GANs for training and calculate the loss. The function should return a tuple of (discriminator loss, generator loss). Use the following functions you implemented:

discriminator(images, reuse=False)
generator(z, out_channel_dim, is_train=True)

def model_loss(input_real, input_z, out_channel_dim):
    """
    Get the loss for the discriminator and generator
    :param input_real: Images from the real dataset
    :param input_z: Z input
    :param out_channel_dim: The number of channels in the output image
    :return: A tuple of (discriminator loss, generator loss)
    """

    # TODO: Implement Function
    
    g_model = generator(input_z, out_channel_dim, is_train=True)
    d_model_real, d_logits_real = discriminator(input_real, reuse=False)
    d_model_fake, d_logits_fake = discriminator(g_model, reuse=True)

    def get_loss(logits, labels):
        return tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(logits=logits, 
                                                                      labels=labels))

    d_loss_real = get_loss(d_logits_real, tf.ones_like(d_model_real))
    d_loss_fake = get_loss(d_logits_fake, tf.zeros_like(d_model_fake))
    g_loss = get_loss(d_logits_fake, tf.ones_like(d_model_fake))

    d_loss = d_loss_real + d_loss_fake

    return (d_loss, g_loss)


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_loss(model_loss)

Tests Passed

Optimization¶

Implement model_opt to create the optimization operations for the GANs. Use tf.trainable_variables to get all the trainable variables. Filter the variables with names that are in the discriminator and generator scope names. The function should return a tuple of (discriminator training operation, generator training operation).

def model_opt(d_loss, g_loss, learning_rate, beta1):
    """
    Get optimization operations
    :param d_loss: Discriminator loss Tensor
    :param g_loss: Generator loss Tensor
    :param learning_rate: Learning Rate Placeholder
    :param beta1: The exponential decay rate for the 1st moment in the optimizer
    :return: A tuple of (discriminator training operation, generator training operation)
    """
    
    # TODO: Implement Function
    
    # Get weights and bias to update
    t_vars = tf.trainable_variables()
    d_vars = [var for var in t_vars if var.name.startswith('discriminator')]
    g_vars = [var for var in t_vars if var.name.startswith('generator')]

    # Optimize
    with tf.control_dependencies(tf.get_collection(tf.GraphKeys.UPDATE_OPS)):
        d_train_opt = tf.train.AdamOptimizer(learning_rate, beta1=beta1)\
                        .minimize(d_loss, var_list=d_vars)
        g_train_opt = tf.train.AdamOptimizer(learning_rate, beta1=beta1)\
                        .minimize(g_loss, var_list=g_vars)    

    return (d_train_opt, g_train_opt)


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_opt(model_opt, tf)

Tests Passed

Neural Network Training¶

Show Output¶

Use this function to show the current output of the generator during training. It will help you determine how well the GANs is training.

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import numpy as np

def show_generator_output(sess, n_images, input_z, out_channel_dim, image_mode):
    """
    Show example output for the generator
    :param sess: TensorFlow session
    :param n_images: Number of Images to display
    :param input_z: Input Z Tensor
    :param out_channel_dim: The number of channels in the output image
    :param image_mode: The mode to use for images ("RGB" or "L")
    """
    cmap = None if image_mode == 'RGB' else 'gray'
    z_dim = input_z.get_shape().as_list()[-1]
    example_z = np.random.uniform(-1, 1, size=[n_images, z_dim])

    samples = sess.run(
        generator(input_z, out_channel_dim, False),
        feed_dict={input_z: example_z})

    images_grid = helper.images_square_grid(samples, image_mode)
    pyplot.imshow(images_grid, cmap=cmap)
    pyplot.show()

Train¶

Implement train to build and train the GANs. Use the following functions you implemented:

model_inputs(image_width, image_height, image_channels, z_dim)
model_loss(input_real, input_z, out_channel_dim)
model_opt(d_loss, g_loss, learning_rate, beta1)

Use the show_generator_output to show generator output while you train. Running show_generator_output for every batch will drastically increase training time and increase the size of the notebook. It's recommended to print the generator output every 100 batches.

def train(epoch_count, batch_size, z_dim, learning_rate, beta1, get_batches, data_shape, data_image_mode):
    """
    Train the GAN
    :param epoch_count: Number of epochs
    :param batch_size: Batch Size
    :param z_dim: Z dimension
    :param learning_rate: Learning Rate
    :param beta1: The exponential decay rate for the 1st moment in the optimizer
    :param get_batches: Function to get batches
    :param data_shape: Shape of the data
    :param data_image_mode: The image mode to use for images ("RGB" or "L")
    """
     
    # TODO: Build Model    
    input_real, input_z, lr = model_inputs(data_shape[1], data_shape[2], data_shape[3], z_dim)
    d_loss, g_loss = model_loss(input_real, input_z, data_shape[3])
    d_opt, g_opt = model_opt(d_loss, g_loss, lr, beta1)
    
    step = 0
    print_every = 10
    show_every = 100
    n_images_to_show = 25

    d_losses, g_losses = [], []
    
    with tf.Session() as sess:
        sess.run(tf.global_variables_initializer())
        
        for epoch_i in range(epoch_count):          
            for batch_images in get_batches(batch_size):
                # TODO: Train Model
                batch_images *= 2
                step += 1

                # Sample random noise for G
                batch_z = np.random.uniform(-1, 1, size=(batch_size, z_dim))

                # Run optimizers
                _ = sess.run(d_opt, feed_dict={input_real: batch_images,
                                               input_z: batch_z,
                                               lr: learning_rate})
                _ = sess.run(g_opt, feed_dict={input_real: batch_images,
                                               input_z: batch_z,
                                               lr: learning_rate})

                if step % print_every == 0:
                    train_loss_d = d_loss.eval({input_real: batch_images,
                                                input_z: batch_z})
                    train_loss_g = g_loss.eval({input_z: batch_z})
                    
                    print_losses(epoch_count, epoch_i, step,
                                 train_loss_d, train_loss_g, 
                                 d_losses, g_losses)

                if step % show_every == 0:
                    print_cumulative_average_losses(train_loss_d, train_loss_g, 
                                                    d_losses, g_losses)

                    show_generator_output(sess, n_images_to_show, input_z, 
                                          data_shape[3], data_image_mode)

                    
def print_losses(epochs, epoch_i, step, train_loss_d, train_loss_g, d_losses, g_losses):
    d_losses.append(train_loss_d)
    g_losses.append(train_loss_g)

    print("Epoch {}/{}...".format(epoch_i+1, epochs),
          "Step: {}...".format(step),
          "Discriminator Loss: {:.4f}...".format(train_loss_d),
          "Generator Loss: {:.4f}".format(train_loss_g))

def print_cumulative_average_losses(train_loss_d, train_loss_g, d_losses, g_losses):
    cum_avg_d_loss = np.mean(d_losses)
    cum_avg_g_loss = np.mean(g_losses)

    print("Cumulative Average Generator Loss / Discriminator Loss: {}".format(
        cum_avg_g_loss / cum_avg_d_loss))
    print("Current Generator Loss / Discriminator Loss: {}".format(
        train_loss_g / train_loss_d))

MNIST¶

Test your GANs architecture on MNIST. After 2 epochs, the GANs should be able to generate images that look like handwritten digits. Make sure the loss of the generator is lower than the loss of the discriminator or close to 0.

batch_size = 16
z_dim = 100
learning_rate = 0.0002
beta1 = 0.2


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 2

mnist_dataset = helper.Dataset('mnist', glob(os.path.join(data_dir, 'mnist/*.jpg')))
with tf.Graph().as_default():
    train(epochs, batch_size, z_dim, learning_rate, beta1, mnist_dataset.get_batches,
          mnist_dataset.shape, mnist_dataset.image_mode)

Epoch 1/2... Step: 10... Discriminator Loss: 0.7566... Generator Loss: 0.7125
Epoch 1/2... Step: 20... Discriminator Loss: 0.1849... Generator Loss: 2.0104
Epoch 1/2... Step: 30... Discriminator Loss: 0.3883... Generator Loss: 2.1976
Epoch 1/2... Step: 40... Discriminator Loss: 2.0035... Generator Loss: 2.5510
Epoch 1/2... Step: 50... Discriminator Loss: 1.5302... Generator Loss: 1.5461
Epoch 1/2... Step: 60... Discriminator Loss: 0.9658... Generator Loss: 1.2796
Epoch 1/2... Step: 70... Discriminator Loss: 1.1134... Generator Loss: 0.5598
Epoch 1/2... Step: 80... Discriminator Loss: 0.6761... Generator Loss: 0.9017
Epoch 1/2... Step: 90... Discriminator Loss: 2.2566... Generator Loss: 0.1425
Epoch 1/2... Step: 100... Discriminator Loss: 1.2161... Generator Loss: 0.9444
Cumulative Average Generator Loss / Discriminator Loss: 1.1581408977508545
Current Generator Loss / Discriminator Loss: 0.7765234112739563

Epoch 1/2... Step: 110... Discriminator Loss: 2.0556... Generator Loss: 0.1852
Epoch 1/2... Step: 120... Discriminator Loss: 1.2008... Generator Loss: 0.6812
Epoch 1/2... Step: 130... Discriminator Loss: 1.0661... Generator Loss: 1.4962
Epoch 1/2... Step: 140... Discriminator Loss: 1.1877... Generator Loss: 1.0268
Epoch 1/2... Step: 150... Discriminator Loss: 1.2055... Generator Loss: 0.6493
Epoch 1/2... Step: 160... Discriminator Loss: 0.8863... Generator Loss: 1.3069
Epoch 1/2... Step: 170... Discriminator Loss: 1.1855... Generator Loss: 2.0561
Epoch 1/2... Step: 180... Discriminator Loss: 1.2612... Generator Loss: 0.5728
Epoch 1/2... Step: 190... Discriminator Loss: 1.0697... Generator Loss: 0.6544
Epoch 1/2... Step: 200... Discriminator Loss: 1.7619... Generator Loss: 0.2607
Cumulative Average Generator Loss / Discriminator Loss: 0.9066960215568542
Current Generator Loss / Discriminator Loss: 0.1479642540216446

Epoch 1/2... Step: 210... Discriminator Loss: 1.8125... Generator Loss: 0.2177
Epoch 1/2... Step: 220... Discriminator Loss: 1.6783... Generator Loss: 0.2775
Epoch 1/2... Step: 230... Discriminator Loss: 1.3983... Generator Loss: 0.4454
Epoch 1/2... Step: 240... Discriminator Loss: 2.0200... Generator Loss: 0.1715
Epoch 1/2... Step: 250... Discriminator Loss: 1.4753... Generator Loss: 0.3042
Epoch 1/2... Step: 260... Discriminator Loss: 2.1541... Generator Loss: 0.1507
Epoch 1/2... Step: 270... Discriminator Loss: 1.0853... Generator Loss: 0.5773
Epoch 1/2... Step: 280... Discriminator Loss: 0.9301... Generator Loss: 0.9153
Epoch 1/2... Step: 290... Discriminator Loss: 0.8079... Generator Loss: 0.9484
Epoch 1/2... Step: 300... Discriminator Loss: 1.0394... Generator Loss: 0.5659
Cumulative Average Generator Loss / Discriminator Loss: 0.6856170892715454
Current Generator Loss / Discriminator Loss: 0.5444575548171997

Epoch 1/2... Step: 310... Discriminator Loss: 0.7195... Generator Loss: 1.5275
Epoch 1/2... Step: 320... Discriminator Loss: 2.0424... Generator Loss: 0.1856
Epoch 1/2... Step: 330... Discriminator Loss: 2.0521... Generator Loss: 0.1564
Epoch 1/2... Step: 340... Discriminator Loss: 2.3246... Generator Loss: 0.1385
Epoch 1/2... Step: 350... Discriminator Loss: 1.5172... Generator Loss: 0.3452
Epoch 1/2... Step: 360... Discriminator Loss: 0.7397... Generator Loss: 1.7611
Epoch 1/2... Step: 370... Discriminator Loss: 1.8919... Generator Loss: 0.2322
Epoch 1/2... Step: 380... Discriminator Loss: 1.0758... Generator Loss: 0.6334
Epoch 1/2... Step: 390... Discriminator Loss: 1.2521... Generator Loss: 0.4407
Epoch 1/2... Step: 400... Discriminator Loss: 0.7564... Generator Loss: 1.2041
Cumulative Average Generator Loss / Discriminator Loss: 0.6244019865989685
Current Generator Loss / Discriminator Loss: 1.5918523073196411

Epoch 1/2... Step: 410... Discriminator Loss: 1.6942... Generator Loss: 2.6384
Epoch 1/2... Step: 420... Discriminator Loss: 1.5494... Generator Loss: 0.2779
Epoch 1/2... Step: 430... Discriminator Loss: 0.6509... Generator Loss: 1.8868
Epoch 1/2... Step: 440... Discriminator Loss: 1.1597... Generator Loss: 0.5298
Epoch 1/2... Step: 450... Discriminator Loss: 1.5094... Generator Loss: 0.3497
Epoch 1/2... Step: 460... Discriminator Loss: 1.5785... Generator Loss: 0.3028
Epoch 1/2... Step: 470... Discriminator Loss: 1.2381... Generator Loss: 0.4447
Epoch 1/2... Step: 480... Discriminator Loss: 0.7624... Generator Loss: 1.4292
Epoch 1/2... Step: 490... Discriminator Loss: 1.1167... Generator Loss: 0.5667
Epoch 1/2... Step: 500... Discriminator Loss: 1.3441... Generator Loss: 0.5348
Cumulative Average Generator Loss / Discriminator Loss: 0.6411036252975464
Current Generator Loss / Discriminator Loss: 0.39791741967201233

Epoch 1/2... Step: 510... Discriminator Loss: 0.9242... Generator Loss: 0.7428
Epoch 1/2... Step: 520... Discriminator Loss: 1.4328... Generator Loss: 3.0068
Epoch 1/2... Step: 530... Discriminator Loss: 0.7721... Generator Loss: 1.3824
Epoch 1/2... Step: 540... Discriminator Loss: 0.9907... Generator Loss: 2.5232
Epoch 1/2... Step: 550... Discriminator Loss: 1.0011... Generator Loss: 0.6148
Epoch 1/2... Step: 560... Discriminator Loss: 1.9046... Generator Loss: 0.2207
Epoch 1/2... Step: 570... Discriminator Loss: 0.9920... Generator Loss: 1.2212
Epoch 1/2... Step: 580... Discriminator Loss: 1.2078... Generator Loss: 0.4195
Epoch 1/2... Step: 590... Discriminator Loss: 1.2881... Generator Loss: 0.4144
Epoch 1/2... Step: 600... Discriminator Loss: 1.6577... Generator Loss: 0.3037
Cumulative Average Generator Loss / Discriminator Loss: 0.6804041266441345
Current Generator Loss / Discriminator Loss: 0.18321171402931213

Epoch 1/2... Step: 610... Discriminator Loss: 1.5464... Generator Loss: 0.2768
Epoch 1/2... Step: 620... Discriminator Loss: 1.1835... Generator Loss: 0.4987
Epoch 1/2... Step: 630... Discriminator Loss: 1.5030... Generator Loss: 0.3097
Epoch 1/2... Step: 640... Discriminator Loss: 1.5632... Generator Loss: 0.3002
Epoch 1/2... Step: 650... Discriminator Loss: 0.9604... Generator Loss: 1.4325
Epoch 1/2... Step: 660... Discriminator Loss: 1.2245... Generator Loss: 1.2434
Epoch 1/2... Step: 670... Discriminator Loss: 1.1483... Generator Loss: 0.6581
Epoch 1/2... Step: 680... Discriminator Loss: 2.3422... Generator Loss: 0.1269
Epoch 1/2... Step: 690... Discriminator Loss: 1.1635... Generator Loss: 0.6014
Epoch 1/2... Step: 700... Discriminator Loss: 1.2318... Generator Loss: 0.4912
Cumulative Average Generator Loss / Discriminator Loss: 0.6421484351158142
Current Generator Loss / Discriminator Loss: 0.3987612724304199

Epoch 1/2... Step: 710... Discriminator Loss: 1.3859... Generator Loss: 0.3754
Epoch 1/2... Step: 720... Discriminator Loss: 1.3893... Generator Loss: 0.3982
Epoch 1/2... Step: 730... Discriminator Loss: 1.2500... Generator Loss: 0.4300
Epoch 1/2... Step: 740... Discriminator Loss: 1.8807... Generator Loss: 0.1882
Epoch 1/2... Step: 750... Discriminator Loss: 0.9911... Generator Loss: 1.6598
Epoch 1/2... Step: 760... Discriminator Loss: 0.9239... Generator Loss: 0.8198
Epoch 1/2... Step: 770... Discriminator Loss: 1.5051... Generator Loss: 0.3305
Epoch 1/2... Step: 780... Discriminator Loss: 0.9147... Generator Loss: 1.4401
Epoch 1/2... Step: 790... Discriminator Loss: 0.9223... Generator Loss: 0.6565
Epoch 1/2... Step: 800... Discriminator Loss: 0.9755... Generator Loss: 1.1815
Cumulative Average Generator Loss / Discriminator Loss: 0.6391062140464783
Current Generator Loss / Discriminator Loss: 1.211110234260559

Epoch 1/2... Step: 810... Discriminator Loss: 1.7624... Generator Loss: 0.2370
Epoch 1/2... Step: 820... Discriminator Loss: 0.7586... Generator Loss: 0.9662
Epoch 1/2... Step: 830... Discriminator Loss: 1.0087... Generator Loss: 0.6304
Epoch 1/2... Step: 840... Discriminator Loss: 0.9913... Generator Loss: 0.8406
Epoch 1/2... Step: 850... Discriminator Loss: 2.4583... Generator Loss: 0.1110
Epoch 1/2... Step: 860... Discriminator Loss: 0.9910... Generator Loss: 0.6721
Epoch 1/2... Step: 870... Discriminator Loss: 1.2124... Generator Loss: 0.4177
Epoch 1/2... Step: 880... Discriminator Loss: 1.2880... Generator Loss: 0.4513
Epoch 1/2... Step: 890... Discriminator Loss: 0.8817... Generator Loss: 0.9843
Epoch 1/2... Step: 900... Discriminator Loss: 0.8398... Generator Loss: 1.8493
Cumulative Average Generator Loss / Discriminator Loss: 0.6336433291435242
Current Generator Loss / Discriminator Loss: 2.201986312866211

Epoch 1/2... Step: 910... Discriminator Loss: 0.8032... Generator Loss: 0.8497
Epoch 1/2... Step: 920... Discriminator Loss: 0.7086... Generator Loss: 1.5629
Epoch 1/2... Step: 930... Discriminator Loss: 1.2632... Generator Loss: 0.5103
Epoch 1/2... Step: 940... Discriminator Loss: 0.8794... Generator Loss: 0.8220
Epoch 1/2... Step: 950... Discriminator Loss: 0.8949... Generator Loss: 2.6132
Epoch 1/2... Step: 960... Discriminator Loss: 1.3966... Generator Loss: 0.3936
Epoch 1/2... Step: 970... Discriminator Loss: 0.6990... Generator Loss: 1.2960
Epoch 1/2... Step: 980... Discriminator Loss: 0.6722... Generator Loss: 1.0808
Epoch 1/2... Step: 990... Discriminator Loss: 1.4927... Generator Loss: 0.3404
Epoch 1/2... Step: 1000... Discriminator Loss: 0.7792... Generator Loss: 0.8229
Cumulative Average Generator Loss / Discriminator Loss: 0.6672875881195068
Current Generator Loss / Discriminator Loss: 1.056157112121582

CelebA¶

Run your GANs on CelebA. It will take around 20 minutes on the average GPU to run one epoch. You can run the whole epoch or stop when it starts to generate realistic faces.

batch_size = 16
z_dim = 100
learning_rate = 0.0002
beta1 = 0.2


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 1

celeba_dataset = helper.Dataset('celeba', glob(os.path.join(data_dir, 'img_align_celeba/*.jpg')))
with tf.Graph().as_default():
    train(epochs, batch_size, z_dim, learning_rate, beta1, celeba_dataset.get_batches,
          celeba_dataset.shape, celeba_dataset.image_mode)

Epoch 1/1... Step: 10... Discriminator Loss: 2.2107... Generator Loss: 0.1456
Epoch 1/1... Step: 20... Discriminator Loss: 0.2086... Generator Loss: 5.2082
Epoch 1/1... Step: 30... Discriminator Loss: 0.2317... Generator Loss: 3.0020
Epoch 1/1... Step: 40... Discriminator Loss: 0.3394... Generator Loss: 1.5442
Epoch 1/1... Step: 50... Discriminator Loss: 0.6039... Generator Loss: 2.6940
Epoch 1/1... Step: 60... Discriminator Loss: 2.5616... Generator Loss: 0.0924
Epoch 1/1... Step: 70... Discriminator Loss: 0.6792... Generator Loss: 0.8867
Epoch 1/1... Step: 80... Discriminator Loss: 0.8176... Generator Loss: 0.7046
Epoch 1/1... Step: 90... Discriminator Loss: 0.7784... Generator Loss: 1.5780
Epoch 1/1... Step: 100... Discriminator Loss: 2.3036... Generator Loss: 0.1257
Cumulative Average Generator Loss / Discriminator Loss: 1.4887502193450928
Current Generator Loss / Discriminator Loss: 0.054565321654081345

Epoch 1/1... Step: 110... Discriminator Loss: 0.6603... Generator Loss: 2.1940
Epoch 1/1... Step: 120... Discriminator Loss: 1.3926... Generator Loss: 0.4210
Epoch 1/1... Step: 130... Discriminator Loss: 2.1155... Generator Loss: 0.1838
Epoch 1/1... Step: 140... Discriminator Loss: 1.5968... Generator Loss: 0.3385
Epoch 1/1... Step: 150... Discriminator Loss: 0.6527... Generator Loss: 1.1434
Epoch 1/1... Step: 160... Discriminator Loss: 0.7111... Generator Loss: 1.9781
Epoch 1/1... Step: 170... Discriminator Loss: 1.3746... Generator Loss: 0.3986
Epoch 1/1... Step: 180... Discriminator Loss: 1.4721... Generator Loss: 0.3108
Epoch 1/1... Step: 190... Discriminator Loss: 0.8788... Generator Loss: 0.7339
Epoch 1/1... Step: 200... Discriminator Loss: 1.1227... Generator Loss: 0.6487
Cumulative Average Generator Loss / Discriminator Loss: 1.0713353157043457
Current Generator Loss / Discriminator Loss: 0.5778168439865112

Epoch 1/1... Step: 210... Discriminator Loss: 0.8176... Generator Loss: 0.8747
Epoch 1/1... Step: 220... Discriminator Loss: 1.6247... Generator Loss: 0.2879
Epoch 1/1... Step: 230... Discriminator Loss: 3.0106... Generator Loss: 0.0582
Epoch 1/1... Step: 240... Discriminator Loss: 2.2705... Generator Loss: 0.1863
Epoch 1/1... Step: 250... Discriminator Loss: 1.0050... Generator Loss: 0.8374
Epoch 1/1... Step: 260... Discriminator Loss: 1.7945... Generator Loss: 0.2747
Epoch 1/1... Step: 270... Discriminator Loss: 1.1722... Generator Loss: 0.4866
Epoch 1/1... Step: 280... Discriminator Loss: 1.0687... Generator Loss: 0.6216
Epoch 1/1... Step: 290... Discriminator Loss: 0.7527... Generator Loss: 0.8829
Epoch 1/1... Step: 300... Discriminator Loss: 1.3495... Generator Loss: 0.4749
Cumulative Average Generator Loss / Discriminator Loss: 0.7801746726036072
Current Generator Loss / Discriminator Loss: 0.35193338990211487

Epoch 1/1... Step: 310... Discriminator Loss: 1.4643... Generator Loss: 0.3096
Epoch 1/1... Step: 320... Discriminator Loss: 1.7791... Generator Loss: 0.2178
Epoch 1/1... Step: 330... Discriminator Loss: 1.0597... Generator Loss: 3.7028
Epoch 1/1... Step: 340... Discriminator Loss: 0.7270... Generator Loss: 1.6809
Epoch 1/1... Step: 350... Discriminator Loss: 0.7829... Generator Loss: 1.4632
Epoch 1/1... Step: 360... Discriminator Loss: 1.5188... Generator Loss: 0.3110
Epoch 1/1... Step: 370... Discriminator Loss: 1.3984... Generator Loss: 0.3599
Epoch 1/1... Step: 380... Discriminator Loss: 0.9021... Generator Loss: 1.1215
Epoch 1/1... Step: 390... Discriminator Loss: 2.2874... Generator Loss: 0.1289
Epoch 1/1... Step: 400... Discriminator Loss: 1.3201... Generator Loss: 0.4850
Cumulative Average Generator Loss / Discriminator Loss: 0.7693768739700317
Current Generator Loss / Discriminator Loss: 0.3674219250679016

Epoch 1/1... Step: 410... Discriminator Loss: 1.9321... Generator Loss: 0.2425
Epoch 1/1... Step: 420... Discriminator Loss: 1.5220... Generator Loss: 0.3281
Epoch 1/1... Step: 430... Discriminator Loss: 0.7729... Generator Loss: 1.7528
Epoch 1/1... Step: 440... Discriminator Loss: 1.4628... Generator Loss: 0.3632
Epoch 1/1... Step: 450... Discriminator Loss: 1.6389... Generator Loss: 0.3001
Epoch 1/1... Step: 460... Discriminator Loss: 0.8783... Generator Loss: 1.1056
Epoch 1/1... Step: 470... Discriminator Loss: 1.0665... Generator Loss: 0.8280
Epoch 1/1... Step: 480... Discriminator Loss: 1.0759... Generator Loss: 0.5815
Epoch 1/1... Step: 490... Discriminator Loss: 0.9675... Generator Loss: 0.7411
Epoch 1/1... Step: 500... Discriminator Loss: 1.7245... Generator Loss: 0.2582
Cumulative Average Generator Loss / Discriminator Loss: 0.714059591293335
Current Generator Loss / Discriminator Loss: 0.14973358809947968

Epoch 1/1... Step: 510... Discriminator Loss: 1.2489... Generator Loss: 0.4608
Epoch 1/1... Step: 520... Discriminator Loss: 1.7654... Generator Loss: 0.2575
Epoch 1/1... Step: 530... Discriminator Loss: 1.5602... Generator Loss: 0.3346
Epoch 1/1... Step: 540... Discriminator Loss: 1.0930... Generator Loss: 0.9979
Epoch 1/1... Step: 550... Discriminator Loss: 1.7739... Generator Loss: 0.2524
Epoch 1/1... Step: 560... Discriminator Loss: 1.3940... Generator Loss: 0.4469
Epoch 1/1... Step: 570... Discriminator Loss: 1.5010... Generator Loss: 0.3642
Epoch 1/1... Step: 580... Discriminator Loss: 1.9409... Generator Loss: 0.1805
Epoch 1/1... Step: 590... Discriminator Loss: 1.3207... Generator Loss: 0.5108
Epoch 1/1... Step: 600... Discriminator Loss: 1.2289... Generator Loss: 0.5275
Cumulative Average Generator Loss / Discriminator Loss: 0.6345771551132202
Current Generator Loss / Discriminator Loss: 0.4292212724685669

Epoch 1/1... Step: 610... Discriminator Loss: 1.3799... Generator Loss: 0.3970
Epoch 1/1... Step: 620... Discriminator Loss: 1.4984... Generator Loss: 0.3487
Epoch 1/1... Step: 630... Discriminator Loss: 1.4992... Generator Loss: 0.3005
Epoch 1/1... Step: 640... Discriminator Loss: 1.0959... Generator Loss: 0.6696
Epoch 1/1... Step: 650... Discriminator Loss: 1.1797... Generator Loss: 0.5919
Epoch 1/1... Step: 660... Discriminator Loss: 1.3260... Generator Loss: 0.6078
Epoch 1/1... Step: 670... Discriminator Loss: 1.1936... Generator Loss: 0.6923
Epoch 1/1... Step: 680... Discriminator Loss: 1.1132... Generator Loss: 0.5602
Epoch 1/1... Step: 690... Discriminator Loss: 1.0304... Generator Loss: 0.7309
Epoch 1/1... Step: 700... Discriminator Loss: 1.3819... Generator Loss: 0.5838
Cumulative Average Generator Loss / Discriminator Loss: 0.6063959002494812
Current Generator Loss / Discriminator Loss: 0.4224366843700409

Epoch 1/1... Step: 710... Discriminator Loss: 1.1363... Generator Loss: 0.5758
Epoch 1/1... Step: 720... Discriminator Loss: 1.2161... Generator Loss: 0.5401
Epoch 1/1... Step: 730... Discriminator Loss: 1.2228... Generator Loss: 0.5161
Epoch 1/1... Step: 740... Discriminator Loss: 1.1206... Generator Loss: 0.6669
Epoch 1/1... Step: 750... Discriminator Loss: 0.9844... Generator Loss: 1.0558
Epoch 1/1... Step: 760... Discriminator Loss: 1.2848... Generator Loss: 0.4869
Epoch 1/1... Step: 770... Discriminator Loss: 1.3451... Generator Loss: 0.4470
Epoch 1/1... Step: 780... Discriminator Loss: 1.2368... Generator Loss: 0.4579
Epoch 1/1... Step: 790... Discriminator Loss: 1.0224... Generator Loss: 0.9698
Epoch 1/1... Step: 800... Discriminator Loss: 1.1080... Generator Loss: 0.7028
Cumulative Average Generator Loss / Discriminator Loss: 0.5999724268913269
Current Generator Loss / Discriminator Loss: 0.6342729926109314

Epoch 1/1... Step: 810... Discriminator Loss: 1.6534... Generator Loss: 0.2625
Epoch 1/1... Step: 820... Discriminator Loss: 1.6885... Generator Loss: 0.2437
Epoch 1/1... Step: 830... Discriminator Loss: 1.1726... Generator Loss: 1.1119
Epoch 1/1... Step: 840... Discriminator Loss: 1.4334... Generator Loss: 0.3577
Epoch 1/1... Step: 850... Discriminator Loss: 0.8967... Generator Loss: 0.8821
Epoch 1/1... Step: 860... Discriminator Loss: 1.2855... Generator Loss: 0.4572
Epoch 1/1... Step: 870... Discriminator Loss: 1.1496... Generator Loss: 0.9019
Epoch 1/1... Step: 880... Discriminator Loss: 1.1470... Generator Loss: 0.9436
Epoch 1/1... Step: 890... Discriminator Loss: 1.2314... Generator Loss: 0.4393
Epoch 1/1... Step: 900... Discriminator Loss: 0.9873... Generator Loss: 1.2781
Cumulative Average Generator Loss / Discriminator Loss: 0.5938462018966675
Current Generator Loss / Discriminator Loss: 1.2945281267166138

Epoch 1/1... Step: 910... Discriminator Loss: 1.6226... Generator Loss: 0.2882
Epoch 1/1... Step: 920... Discriminator Loss: 1.8125... Generator Loss: 0.2149
Epoch 1/1... Step: 930... Discriminator Loss: 1.8262... Generator Loss: 0.2369
Epoch 1/1... Step: 940... Discriminator Loss: 1.0133... Generator Loss: 1.3767
Epoch 1/1... Step: 950... Discriminator Loss: 1.9051... Generator Loss: 0.2045
Epoch 1/1... Step: 960... Discriminator Loss: 1.8655... Generator Loss: 0.2231
Epoch 1/1... Step: 970... Discriminator Loss: 2.2053... Generator Loss: 0.1382
Epoch 1/1... Step: 980... Discriminator Loss: 1.2546... Generator Loss: 0.5030
Epoch 1/1... Step: 990... Discriminator Loss: 1.0137... Generator Loss: 0.8261
Epoch 1/1... Step: 1000... Discriminator Loss: 1.3875... Generator Loss: 0.3847
Cumulative Average Generator Loss / Discriminator Loss: 0.5554788112640381
Current Generator Loss / Discriminator Loss: 0.27727654576301575

Submitting This Project¶

When submitting this project, make sure to run all the cells before saving the notebook. Save the notebook file as "dlnd_face_generation.ipynb" and save it as a HTML file under "File" -> "Download as". Include the "helper.py" and "problem_unittests.py" files in your submission.